1. Field of the Invention
The present invention relates to a semiconductor device and a method for manufacturing the semiconductor device.
2. Related Art
Flash memories that are nonvolatile memories, especially NAND flash memories, have been vigorously developed as still-image recording media and high-quality audio recording media, since such flash memories can be readily made smaller in size, and can be provided with a large amount of memory at low costs. Furthermore, such flash memories are shock-resistant. As a result, there is now a big market for flash memories.
Each memory cell used in a NAND flash memory has a gate structure on a semiconductor substrate. In this gate structure, a tunnel insulating film, a charge storage film, an interelectrode insulating film, and a control electrode are stacked in this order. The gate structure may be of a floating gate type (FG type) having a floating gate electrode made of polysilicon to be the charge storage film on the tunnel insulating film, a MONOS (Metal-Oxide-Nitride-Oxide-Silicon) type having a charge trapping film made of silicon nitride to be the charge storage film, or a SONOS (Silicon-Oxide-Nitride-Oxide-Silicon) type having the charge storage film made of a nitride and the control electrode made of silicon.
The threshold voltage of each memory cell can be varied in the following manner. By controlling the voltage (control voltage) to be applied to the control electrode formed on the floating gate or the charge trapping film via the interelectrode insulating film, electrons are injected (written) from the substrate into the floating gate electrode or the charge trapping film via the tunnel insulating film through FN (Fowler-Nordheim) tunneling, or electrons are pulled out of the floating gate electrode via the tunnel insulating film (erasing in the FG, MONOS, or SONOS type). Also, holes are injected into the charge trapping film, so that the holes and electrons annihilate each other (auxiliary erasing in the MONOS/SONOS type). However, as the tunnel insulating film has become thinner in the trend of miniaturization, attention is now drawn to one serious problem.
To achieve a larger amount of memory, reducing the device size (the memory cell size) is the most effective, but the tunnel insulating film needs to be thinner at the same time. SiO2 film that is widely used as the tunnel insulating film characteristically experiences more leakage current particularly in a low-field region, called SILC (stress Induced Leakage Current), via the trapping center existing in the SiO2 film, due to application of a stress voltage as well as the smaller film thickness. As a result, the amount of charges passing through the SiO2 film increases, and the data retention properties are degraded. The time elapsed before the charge amount reaches a destructive value becomes shorter accordingly, or the rewriting performance is degraded. The SILC hinders a reduction of the SiO2 film thickness, and deteriorates the reliability, resulting in great difficulties in miniaturization. Unless this trapping center is reduced, a larger amount of memory cannot be expected by reducing the film thickness of the SiO2 film.
As described above, as a known cause of degradation of the characteristics of the SiO2 film, an interface layer made of SiO2 in an amorphous state is formed between the semiconductor substrate and the SiO2 film, and oxygen defects always exist in the amorphous SiO2. This results in various traps and leakage sites.
Dangling bond of Si at SiO2/Si interface is normally terminated with hydrogen. However, the hydrogen is detached from the dangling bond of Si by electrons or holes during a rewriting operation, and does not provide a fundamental solution. It has been known that terminating a dangling bond of Si with deuterium is effective, but it remains unclear whether the use of deuterium is effective at the SiO2/Si interface.
As a solution to the problem, the following method has been suggested. Nitrogen is introduced into the SiO2 film to be the tunnel insulating film, so as to increase the dielectric constant and the physical film thickness, and reduce the leakage current. However, the effect of this method is not sufficient, and the thinnest possible film thickness is not as thin as expected. This is because the defect formation due to the insufficient Si—N network is not appropriately restricted.
To counter this problem, there has been a structure in which a three-layer structure having a high-quality silicon nitride film interposed between silicon oxide films is used as the tunnel insulating film that does not easily allow defect formation, and the silicon nitride film has three-coordinate nitrogen bonds (see JP-A 2007-059872(KOKAI)). To form the insulating film having the stacked structure formed with a SiO2 layer, a SiN layer, and a SiO2 layer, the following method is known. After a SiO2 layer is formed on a Si substrate, amorphous Si is deposited and is nitrided to form a nitride layer made of SiN. The nitride layer is then oxidized, or an oxide layer is deposited by CVD.
However, as a result of a study made by the inventor, the following problems were found with the above method. 1) If nitridation is performed at a high temperature, the amorphous Si is crystallized and agglomerated to form a grain boundary, and the layer thickness of the Si layer fluctuates. 2) The concentration of hydrogen remaining in the amorphous Si might become high. In the case of the problem 1), if the SiO2 layers are thin, nitrogen penetrates through the lower SiO2 film, and defects are formed at the interface between the SiO2 layer and the Si substrate. When a current flows from the Si substrate to the control electrode, not only the amount of current in the low electric field or the medium electric field increases, but also the insulating properties are severely degraded due to a local decrease in the layer thickness of the nitride layer in a worst case scenario. In the case of the problem 2), the reliability deteriorates.